Do legislators vote their constituents' wallets?

• Author: Bogart, William T.
• Position: And how would we know if they did?

1. Introduction

   Do the economic consequences of policies affect the political behavior of elected officials? In particular, do the costs and benefits that flow to households in legislative districts affect representatives' support of policy proposals? In recent years, scholars have investigated this question by regressing roll-call vote decisions of legislators against measures of the economic interests of constituents and the ideology of the legislators |3; 5; 8; 9; 10; 11; 13; 14~. With some exceptions, such as Peltzman |13~, most scholars who perform such regressions conclude that constituents' economic interests are not the sole determinant of legislators' behavior. Representatives' own policy preferences explain some of the cross-sectional variance in their support of proposals. In the words of economists, representatives shirk from their constituents' interests.

   Many authors who regress roll-call vote behavior against measures of constituents' economic interests and representatives' ideology are aware that their measure of the latter (summary roll-call vote ratings by groups such as the Americans for Democratic Action (ADA)) is not a pure measure of representatives' policy beliefs because it is merely a summary of previous roll-call behavior. To eliminate the contamination, scholars regress the ADA rating against numerous economic, demographic, and regional variables. The unexplained residual variance is then used to measure representatives' ideology.

   Those who use this technique obviously hope that these residuals represent personal ideology and not other possible causes of vote decisions. The reasonableness of this assumption depends on whether the variables regressed against the ADA rating capture all the other causes of congressional behavior. Kalt and Zupan |9, 293; 10, 112~, Carson and Oppenheimer |5, 172~ and Kau and Rubin |11, 64~ exclude three important classes of variables: the positions of interest groups and executive-branch actors, and the likelihood of proposal passage, all of which affect the decisions of legislators. This exclusion increases the probability that the residuals in the ADA equation represent not only the personal ideology of representatives, but also the ability of executive-branch and interest-group actors to affect reelection.(1)

   Even though critics of this line of research focus on the nature of the ADA residuals, the measures of constituents' economic interests are equally problematic. In Peltzman's |13, 184~ words, "The usual procedure in modeling legislators as agents has been to treat all the residents of a legislator's district as his principals. That is, the vector (of exogenous economic characteristics) usually consists of a set of average resident characteristics, that is, per capita income, education, and so on." Individual-level economic data are not used to measure the economic affects of policy proposals or to estimate the political salience of the economic effects.

   In this paper, we do not attempt to ascertain the effects of constituents' economic interests and legislators' ideology on the policy choices of the latter. Instead, we estimate the economic impact of a policy change at the household level and then ask whether the behavior of legislators toward the policy change is consistent with any one of several plausible aggregations of households' economic interests into a political referendum under majority rule. Following Coates and Munger |6~, we also test whether the cross-sectional variation in legislators' tendency to make policy decisions that are in accord with their constituents' economic interests varies with the margin of victory received by the legislator in the previous election. In addition, to the extent that legislative behavior does not correspond to any plausible aggregation of household economic effects, we examine whether voters react in the subsequent election by punishing those legislators who did not vote in their constituents' economic interests.

2. Economic Factors and Legislative Behavior: A New Approach

   In this paper, we examine policy choices in which it is likely that the effect of economic benefits will be salient and we use data that allow us to measure those benefits at the household level. The fiscal reforms enacted by New Jersey in 1990 and 1991 were (and are) extremely salient, and legislators did not use parliamentary tactics to disassociate themselves from their decisions to enact the changes in taxes and aid |2~. If legislators ever respond to the economic effects of policies on constituents, we surely should observe such behavior in circumstances like those experienced in New Jersey in 1990 and 1991.

   We use a unique data set of matched income and property tax data for all New Jersey homeowners to estimate the net costs and benefits at the household level of the system of income taxes, property taxes, property tax rebates, and school aid during 1987. We then estimate for every household the net benefits of the massive fiscal changes enacted by the New Jersey legislature in 1990: a large increase in income taxes, school aid to poorer jurisdictions, and the redistributive character of property tax rebates. We do the same for additional reforms enacted in 1991: a decrease in school aid to the poorest jurisdictions and an increase in general municipal aid. Finally, for both these policy changes, the household data are grouped into legislative districts to determine whether state senators' decisions to support or oppose the reforms enacted in 1990 and their partial repeal in 1991 reflect the economic interests of households calculated either cardinally (the mean voter, dollar-weighted) or ordinally (the median voter).

   These data allow us to directly observe the economic effects of various tax and expenditure policies at the household level and eliminate the need to use the imperfect proxies so often observed in roll-call studies. In addition, these data make clear that the determination of whether the economic interests of constituents have been adequately represented depends on whether the costs and benefits are aggregated across individuals (1) cardinally or ordinally and (2) district by district or statewide.

3. The Incidence of School Aid and Property Tax Rebates

   The incidence of the state and local government tax and spending system is usually approximated by the sum of state and local government taxes. Intergovernmental grants, a form of state government spending, are usually considered in isolation, because the incidence of spending is usually treated separately from the incidence of taxes. However, even though grants are an expenditure item for the state government, they represent an important source of revenue for local governments. Hence, any analysis of a change in grants distribution that does not account for the resulting change in local taxes will be incomplete. It is important, therefore, to consider the role of grants as a local revenue source and their impact on local voters as an increase in local government resources. Aside from this change in the treatment of grants, we follow the rest of the literature on tax incidence and do not explicitly consider the incidence of state or local government spending.

   The government receiving an intergovernmental grant may direct the money either to reducing local taxes, increasing local government spending, or both. In order to identify the effect of the grant on individual taxpayers, one must make some assumptions about the allocation of the grant between these two uses.(2) We assume that reductions in local taxes and increases in local government spending are valued equally and identically distributed among local taxpayers. In other words, we will model the incidence of a change in grants as being the same as the incidence of a change in local taxes. We now turn to the question of determining the incidence of local taxes.

   The standard theory of tax incidence suggests that in an open economy (such as New Jersey and its constituent municipalities)(3), the burden of taxes will be borne by relatively immobile factors of production. The most immobile factor is real estate (sites). For the purposes of our analysis, we treat sites as the only immobile factor, and assume that all gains and losses resulting from the shift in fiscal policy are reflected in site values.(4)

   It is straightforward to implement our incidence assumption using our data on New Jersey homeowners. Every property has an value assigned to it by the local government for tax purposes. Clearly, one can total these assessments over all properties within a jurisdiction. Define an individual's tax share to equal his assessed value divided by the total assessed value in a municipality. Note that the property tax levy and the total assessed value are both determined by the local government. (The tax rate is uniquely defined by this process, as it equals the tax levy divided by the total assessed value.) An individual's property tax payment is simply his tax share multiplied by the total property tax levy. If the total levy falls as a result of intergovernmental grants, then an individual's property tax falls by the amount of his tax share multiplied by the grant. This is the way in which grants are allocated among individuals in the empirical work that follows.(5)

   The incidence considered in this work is the "impact" incidence of intergovernmental grants. The long run incidence of the grants will depend on the reactions of individuals and on whether individuals are able to transfer tax increases to others. The long run incidence of a reform is not...